Instructions for replicating "Predicted Probabilities and Inference
with Multinomial Logit"

The analysis in this paper was conducted using Stata version 15.1 on a
Dell Optiplex 7060 with an Intel(R) Core(TM) i7-8700 CPU @ 3.20GHz
with six cores and 16GiB System Memory running Ubuntu 18.04.  In order
to reproduce all analysis in the paper and the supplemental materials,
copy all of the files in 1. through 4. to a directory and run
mlogit_replication.do using Stata. The runtime for the .do file is
6 seconds.  Users will need the estsimp.ado, setx.ado, and simqi.ado
files from Tomz, Wittenberg, and King's (2003) Clarify program
(https://gking.harvard.edu/clarify) in order to estimate the results
in Gelpi's (2017) Figure 3.

The replication files are as follows:

1. mlogit_replication.do: The stata command file for reproducing all
results in the main paper and supplemental materials.

2. GelpiFeb2009.dta and GelpiPOC.dta: The two data files needed for
reproducing the results for Gelpi's "Performing on Cue" paper and the
additional analysis of that work for this paper.

3. 2015-05-0216.R2 Rumor_Has_It--_Thailand_replication_data.dta and
2015-05-0216.R2 Rumor_Has_It--_Philippines_replication_data.dta: The
two data files needed for reproducing the results for Greenhill and
Oppenheim's "Rumor Has It: The Adoption of Unverified Information in
Conflict Zones" paper and the additional analysis of that work for
this paper.

4. replication data.dta: The data file needed for reproducing the results
for Weber, et al.'s "Placing Racial Stereotypes in Context: Social
Desirability and the Politics of Racial Hostility" paper and the
additional analysis of that work for this paper.

5. mlogit_replication.out: The output file that shows the output from
running the mlogit_replication.do file.

6. tables/table1.txt tables/tableA1.txt tables/tableA2.txt are files
that contain Stata output with the information from which Tables 1,
A.1, and A.2 were created.

7. figures/figure1a.pdf figures/figure1a.pdf figures/figureA1a.pdf
figures/figureA1b.pdf are the figures generated by Stata that are
reproduced in the paper and online supplemental appendix.

Two other files that are not needed for replicating the work in the
paper, but may be of use to some readers are:

gelpi_replication_Rcode.R: R code for generating predicted
probabilities and their standard errors.

gelpi_replication_Rcode.dta: A Stata data file with the recoded
variables from the analysis of Gelpi's "Performing on Cue" paper.
Even though this file is reproduces the analysis, the two files given
above for reproducing Gelpi's work should be referred to because they
contain Gelpi's data prior to new recodes.
